Automated Image Captioning

Abstract

Image captioning means automatically generated a caption or description of a given image. Image captioning required feature detect at the best level. Various methods have been proposed early. We have a retrieval-based and template base. After improving the deep neural network, many deep learning methods are proposed based on different frameworks. To perform image captioning, there exist many datasets with different features and attributes. To find how good the automatic caption is for the generation, we have different evolution metrics. We introduce the data augmentation on caption and image of Flickr8k dataset and evaluate our model on Flickr8k dataset to get the score of BLUE-1 0.59 and BLUE-4 0.20, respectively as compare to without augmentation and get the score of BLUE-1 0.55 and 0.15 respectively.